[Bugfix] Fix BNB loader target_modules

[jeejeelee]

Motivation

When addressing the following TODO, I found a misconception regarding the use of target_modules. The issue arises because target_modules specifies which layers should have LoRA adapters added, rather than which layers should be quantized.

vllm/vllm/model_executor/model_loader/loader.py

Line 714 in 15cc2a9

# TODO: target_modules could be either a list or a regex string.

Reproduce Code

from transformers import AutoModelForCausalLM
from peft import get_peft_model, LoraConfig, TaskType

model_name_or_path = "Llama-2-7b-chat-hf"
peft_config = LoraConfig(
    task_type=TaskType.CAUSAL_LM,
    inference_mode=True,
    r=8,
    lora_alpha=32,
    target_modules=[ # just add some layers to verify the effect of target_modules
        "gate_proj",
        "up_proj",
    ],
    use_rslora=False,
    use_dora=False,
)
model = AutoModelForCausalLM.from_pretrained(
    model_name_or_path, load_in_4bit=True
).eval()
model = get_peft_model(model, peft_config)
model.print_trainable_parameters()
print(model)

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[jeejeelee]

@mgoin Sorry for bother you, if you have bandwidth ,please look at this PR, thanks

[mgoin]

Apologies for missing, this looks reasonable to me if the CI is green!